You Can’t Detect an Improviser: Why Self-Learning AI Malware Ends the Detection Era 

The report warns of a new class of AI-driven "self-learning" malware that adapts and rewrites itself across hosts, citing a proof-of-concept worm and industry observations of rising AI-enabled attacks; it argues that detection-based tooling (signatures, EDR/XDR) will be insufficient against uniquely morphing threats and recommends a prevention-first approach (Automated Moving Target Defense) to block execution before damage occurs.